1

An Active-shunt Diverter for On-load Tap ChangersDaniel J. Rogers, Member, IEEE, and Tim C. Green, Senior Member, IEEE

Abstract—This paper presents a new hybrid diverter designfor on-load tap changers. The design uses ‘active-shunt’ currentdiversion principles. At its core, the design employs a low-voltagehigh-current switch-mode amplifier to divert current out of themechanical contacts and into a pair of anti-parallel thyristors.Commutation between transformer taps may then be performedby the thyristors. The amplifier and thyristors are placed outsidethe normal load current path and only conduct during a tapchange, producing efficiency savings and improving robustnesswhen compared to previous hybrid OLTC implementations. Anamplifier control loop that autonomously produces zero-currentconditions at switch opening and zero-voltage conditions at switchclosure is demonstrated

Experimental results investigating the wear characteristics ofcontacts operated under the new hybrid diverter are presented,along with comparison results from a passive-type switchingscheme. Contact lifetime of over 25 million operations is demon-strated under the new scheme.

I. INTRODUCTION

THE distribution network of the future faces very signifi-cant challenges in accommodating the expected degree of

penetration of distributed renewable energy and electric vehi-cle charging. It has been recognised that export of power fromPV systems causes voltage rise in distribution networks andthat strategies to combat this are needed [1], [2]. Manipulationof reactive power by PV systems and curtailment of powerexport are options [3] but clearly curtailment is economicallyunattractive and reactive power becomes less effective as theX/R ratio reduces in low voltage networks. It is also knownthat PV power export is subject to rapid fluctuations of powerproduction due to the passing of cloud shadows. Studies [4]have shown that considerable smoothing of output is obtainedbetween even adjacent systems in the 10 second time frame butover minutes the variations are correlated and large aggregatefluctuations occur. It was found that a separation of 50 km wasrequired for changes over 30 minutes to be uncorrelated.

Existing distribution network voltage control mechanisms,which rely on on-load tap-changers (OLTCs), were designedfor slow adjustment of gradual demand changes over a dailycycle. It is possible to provide fast and accurate voltage controlusing “electronic transformers” (for instance [5]) but as yetthese devices are not competitive in terms of power losses,capital cost or footprint with conventional OLTC equippedtransformers. The duty of an OLTC providing voltage controlon feeders rich in PV and EV could be very high. As anillustration, suppose that the OLTC reacted at 5 minute inter-vals and that on most occasions only a single-tap increment

This work was funded by a UK EPSRC Doctoral Training Award and wasconducted partly within the Supergen FlexNet programme.

D. J. Rogers (email: RogersDJ@cardiff.ac.uk) is with the Insti-tute of Energy, Cardiff University, Cardiff, UK. T. C. Green (email:t.green@imperial.ac.uk) is with the Electrical Engineering Department, Impe-rial College London, London, UK.

was needed but on 10 % of occasions 10-tap increments wererequired due to either shadowing of PV or correlated EVcharging. It is possible that 300 tap changes per day maybe seen, leading to 110,000 tap operations per year and 4.5million tap operations over a 40 year life. OLTC manufacturersare now offering devices based on vacuum switches ratherthan more traditional contacts in oil [6], [7]. These are quotedas having a service interval of 300,000 operations and alifetime of 2 million operations [8]. For widespread use indistribution substations of the future, much greater serviceintervals are needed and it is probably necessary to reducethe physical volume of the OLTC diverter element to allowretro-fit to existing substations. This observation provided themotivation for the work to be described here which seeks toachieve wear-less operation of OLTC contacts by providingzero-current opening and zero-voltage closing with a shunt-connected active element.

Whilst yet to see significant commercial uptake, severalOLTC designs containing semiconductor devices have beenpresented in the literature in the past fifty years. These designsseek to dramatically reduce the degree of arcing occurringat the mechanical diverter contacts. Although some designsdirectly replace the mechanical contacts with semiconductordevices (e.g. [9], [10]) it is often suggested that mechanicalswitching elements should not be eliminated entirely. Hybrid(or ‘semiconductor-assisted’) OLTC designs are presented inwhich semiconductor devices are used to perform currentinterruption/commutation but mechanical switching elementsare retained for use during steady-state operation. In this way,the strengths of both devices are exploited: semiconductordevices provide wear-less commutation capability during atap change operation whilst mechanical contacts provide verylow on-state resistance and correspondingly small power lossduring steady-state (fixed tap) operation. The method by whichcurrent transfer between mechanical switch and semiconductordevice is achieved during the commutation process leads toa natural separation of the hybrid diverter designs into threecategories, as shown in Fig. 1.

The objective of this paper is to explore in detail thecircuit of 1(c) (the active-shunt diverter) in comparison withthose of 1(a) and (b) (the passive and active-series diverters).The active-shunt circuit was introduced in [11], [12] andexperimental results used to demonstrate its effectiveness. Inthis paper the tap-sequence and operation of the active elementare described in detail (with an improved control configurationover [11]), a discussion of arc energy and its implication forcontact wear is given, and a full comparison of the passiveand active configurations is undertaken. To begin, these threediverter configurations will be described.

2

A. Passive type

The passive-type hybrid OLTC relies on the process ofmechanical contact separation to drive current into an alternatepath. Typically, the alternate path contains a pair of anti-parallel thyristors (either line commutated or gate-turn offtype). An example passive scheme is that of [13] in whichthe increasing voltage drop across an opening mechanicalcontact is used to trigger a thyristor into conduction via apulse transformer, providing an alternate current path at someinstant after the moment of contact separation. The principledrawback of passive-type schemes is that arcing at the switchcontacts may never be entirely eliminated. Fig. 1(a) illustratesa conceptual circuit that includes two parasitic inductances(labelled LS and LT ) that inevitably occur in any practicalcircuit implementation. Upon opening of the switch, andassuming the forward thyristor is triggered into conductionbeforehand, these inductances act to limit the rate of transferof load current (IAC) between the switch (IS) and the thyristor(IT ) paths.

B. Active-series type

Active-series hybrid OLTC designs are characterised by theplacement of an opposing voltage source (VD) in series withthe mechanical switch. The activation of this voltage sourcecauses current to transfer out of the switch path and intoan alternative path containing semiconductor devices such asthyristors (see Fig. 1(b)). This is achieved without opening theswitch under load: after the current has been fully transferredthe switch may be opened under a condition of zero current,eliminating contact arcing. An example of an active-seriesdesign is found in [14] where the series voltage source istermed the ‘auxiliary current diverter’.

C. Active-shunt type

In the active-shunt configuration illustrated in Fig. 1(c), thevoltage source is placed in the thyristor diversion path insteadof the switch path. The source is connected such that it actsto cancel the thyristor forward voltage and is controlled todrive a current equal to the load current through the alternatepath (IT = IAC). In this case the current in the switch will bezero, and arc-less operation of the switch is ensured. The mainadvantage of the active-shunt scheme over the active-seriesscheme is that no additional device must be inserted in serieswith the steady-state load current path (i.e. the switch path).Thus, no additional steady-state power loss is incurred and,for the majority of time, no semiconductor device is exposedto the load current and hence to possible fault currents. Onlyduring current diversion (i.e. during a tap change operation)is the active device required to conduct the load current. Thedisadvantage of the active-shunt design is that active control ofthe voltage source is required to provide sustained zero-currentconditions in the switch.

In a well-designed active system the current at switch open-ing will be very low, governed by leakage current through thevoltage source (for the series design) or by control-loop errors(for the shunt design). Therefore switch contacts operated

under either scheme may be expected to experience low wearrates and deliver very long service lifetimes. The primaryoperational difference lies in the fact that the shunt design doesnot experience the steady-state power loss associated with thecontinually conducting voltage source of the series design.

In both the active-series and active-shunt circuits the voltagesource, whilst required to conduct the full load current, is onlyrequired to sustain a voltage approximately equal to that of theconducting thyristor in order to cause current diversion. ForLV and MV OLTC systems, single-junction thyristors provideample voltage blocking capability and the required sourcevoltage will less than 10 V.

LSLT

IAC

ISIT

VSVT

(a) Passive

LSLT

IAC

ISIT

VSVT

VD

(b) Active-series

LSLT

IAC

ISIT

VSVT

VD

(c) Active-shunt

Fig. 1. Classification of hybrid OLTC schemes

II. AN ARC-LESS OLTC BASED ON THE ACTIVE-SHUNTPRINCIPLE

Fig. 2 shows the high-level circuit implementation and theoperating steps of a new OLTC diverter based upon the active-shunt principle. Similar to most OLTCs, the design can beseparated into a selector section and a diverter section. Theselector section is capable of connecting any odd-numberedtap to the left ‘leg’, and any even-numbered tap to the right leg.The system operates by transferring the load current betweenlegs; each transition allows a different tap to be selected, aslong as the new tap is on the opposing leg. The selectorcontacts are never required to make or break current; thisfunction is handled by the diverter section. The diverter sectionof the new OLTC design consists of two mechanical switches(SL and SR), two sets of anti-parallel thyristors ([TL+, TL−]and [TR+, TR−]), and a ‘controlled source’ which may beoperated such that it behaves as either a current source or avoltage source as required (hence it is illustrated as both inFig. 2).

The fundamental task of the diverter sub-circuit in the newdesign is identical to that of the diverter in the classic, purelymechanical OLTC: to transfer the load current between theswitches without interruption. However, an additional task ofthe new diverter is to ensure zero-current or zero-voltage

3

SS4

SS2

SR

SS3

SS1

SL

IAC

TL-

TL+

TR-

TR+

off

VAC

ZLOAD

diverter

selector

(a) t < tC . Steady-state operation ontap 1. Controlled source off.

SS4

SS2

SR

SS3

SS1

SL

IAC

TL-

TL+

TR-

TR+

on

VAC

ZLOAD

(b) tC < t < tF . SL is opensunder a condition of zero current.SS2 begins closing in preparationfor operation on tap 2.

SS4

SS2

SR

SS3

SS1

SL

IAC

TL-

TL+

TR-

TR+

on

VAC

ZLOAD

(c) tF < t < tI . SL open. SR closesunder a condition of zero voltage.SS1 begins opening in preparation foroperation on tap 2.

SS4

SS2

SR

SS3

SS1

SL

IAC

TL-

TL+

TR-

TR+

off

VAC

ZLOAD

(d) tF < t < tG. Steady-stateoperation on tap 2. SR is closed andthe controlled source is off.

Fig. 2. Current paths in the active-shunt OLTC design when performing asingle tap-up operation from tap 1 to tap 2. See Fig. 3 for time references. Afiner-grained breakdown of the tap-change sequence is given in [11].

conditions are maintained during switch operation. Whileeither the left or the right switch is closed, a condition ofzero current may be created by triggering the correspondingthyristor pair into conduction and operating the controlledsource in current mode such that it produces a current equal tothe load current. In this case, the load current flows through thecontrolled source leaving zero residual current in the switch.Similarly, during a period in which either switch is open, acondition of zero voltage may be created by triggering thethyristor pair into conduction and operating the controlledsource in voltage mode such that it produces a voltage thatexactly counterbalances that of the conducting thyristor. Inthis case, the sum of voltages in the loop around the switchis zero and there is zero potential across the switch contacts.

Commutation between left and right legs of the diverteris performed at the zero-crossing of the load current. Thetrigger signal for the outgoing thyristor is removed so thatit transitions to the blocking state at the zero-crossing, atwhich point the reverse-facing thyristor on the opposing leg istriggered. Fig. 3 shows the timing of the transitions betweensteps.

A. The diverter circuit and the controlled source

Fig. 4 shows one possible device-level implementation ofthe hybrid-shunt diverter. The controlled source is switch modein nature, employing four low-voltage, high-current MOSFET

V I I V V→II→V

SS1

SS2

TL+

TR-

ID / VD

IAC

SL

SR

tAt

tB tC tF tI tKtJtD tHtE tG

Fig. 3. Idealised timing diagram for a tap change performed over onecycle of the load current. A high level for SS1, SS2, SL, SR indicatesthe corresponding switch is closed, a low level that the switch is open,intermediate values represent a switch that is changing state and is neitheropen or closed. A high level for TL+, TL− indicates that trigger currentis applied. A high level for ID /VD indicates that the controlled sourceis operating, the I or V entry denotes current or voltage source moderespectively. Shaded regions indicate device conduction.

devices in an H-bridge configuration. The inductor LD isincluded so that current-source behaviour may be producedthrough suitable closed-loop control of the H-bridge duty cyclewhen one of the two switches (SL or SR) is closed. CD andRD are present to define the voltage VD when both switchesare open. RC snubbers are placed across both thyristor pairsto aid commutation between diverter legs at the load currentzero crossing.

The H-bridge must be capable of driving a sinusoidalcurrent equal to the load current through the inductor LD(ILD). The maximum average value of VA is equal to thesupply voltage VDC . Given a worst case thyristor forwardvoltage drop of VT (max) and assuming negligible parasiticseries resistances the maximum allowable size of LD is givenby

LD(max) =VDC − VT (max)

ω0I0, (1)

where I0 is the peak load current and ω0 is the line angularfrequency. For cost and size reasons it is desirable to keepLD and VDC small. However, the design must be tolerant ofvariations in the inductor value, as well as parasitic circuitresistances that will tend to reduce the achievable rate ofchange of current through the inductor. Therefore reasonablechoices are VDC = 2VT (max) and LD = LD(max)/2. Fordistribution network applications VDC and LD will lie inthe region of 7 V and 50µH respectively. CD and RD areless critical in terms of their sizing, but they should bechosen such that they provide good attenuation of the ripplecurrent produced by the H-bridge. For distribution networkapplications, CD and RD will be in the region of 100µF and1 Ω respectively.

The use of MOSFETs in an OLTC represents an interestingthird level of hybridisation over that seen in previous systems.Mechanical switches are exploited for their low conductionlosses and high robustness, thyristors are used for their high-

4

SRSL

TL-

TL+

TR-

TR+

CD

LD

RD

CNL RNL RNR CNR

IAC

ID

controlled source

IS

VDC

VIT

VDVSRVSL

VTL VTR

ICDILD

Fig. 4. The active-shunt OLTC diverter circuit

voltage high-current wear-less commutation capabilities, andMOSFET devices are included to provide a highly control-lable, low-voltage high-current amplifier device in the form ofthe controlled source.

B. Commutation between diverter legs

Current commutation between diverter legs occurs aroundthe zero-crossing of the line current (e.g. between Figs. 2(c)and (d)). If the commutation process is not handled carefullyan inter-tap short-circuit may be established should the in-coming thyristor be triggered before the outgoing thyristor hasfully recovered its blocking capability. For this to occur, theinter-tap voltage VIT must be of a magnitude and directionthat will forward bias the incoming and outgoing thyristors.This scenario is encountered in exactly half of all possiblecommutation events. A commutation event is described bythe following four variables: tap direction (up or down),commutation direction (left-to-right or right-to-left), currentzero-crossing direction (positive-to-negative or negative-to-positive) and power factor (load current leading or laggingthe inter-tap voltage). An exhaustive list of potential short-circuit cases may be generated by considering all possiblecombinations of the these variables. In order to avoid theintroduction of an inter-tap short-circuit, the triggering of theincoming thyristor is delayed for a short time after the currentzero-crossing in order to guarantee that the outgoing thyristorhas fully recovered its forward blocking capability. In thisperiod the load current must flow in the snubber components,the resulting snubber voltage should be limited by appropriateselection of the snubber components (CNL, RNL, CNR andRNR).

C. A diverter feedback law

Fig. 5 shows one possible feedback law arrangement thatis capable of producing the desired zero-current zero-voltageconditions. The system requires measurements of the sumof the switch currents (IS), and both switch voltages (VSL

and VSR). The diverter circuit exhibits different behaviourdepending on whether one of the switches (SL or SR) areclosed or both are open (the switches may never be closedat the same time as this would introduce an inter-tap short-circuit). This implies that two different plant models arerequired when designing the compensators G(s) and H(s),one that represents the circuit with one switch in the closedstate and one that represents the circuit with both switchesin the open state. Simplified control loop models for bothstates are illustrated in Fig. 6. Whilst operating with a switchclosed, the corresponding voltage measurement across thatswitch is by definition zero, hence the output of the H(s)block is also zero and G(s) acts to drive IS to zero. Thisis termed the current mode. Similarly, when both switchesare open, IS is by definition zero and therefore the outputof the G(s) block is zero; H(s) acts to drive either VSL orVSR to zero depending on the position of the selector switch.This is termed the voltage mode. The effect of the (non-linear)thyristor forward voltage (VTL or VTR in Fig. 4) is modelledby the disturbance input Vdis, this voltage may be consideredalmost independent of the control loop action during normaloperation due to the weak dependence of thyristor voltage onforward current. While in current mode, the rapid reversal ofthis voltage at the current zero-crossing acts to drive a currentimpulse through CD (ICD in Fig. 4) limited in magnitude byRD. This current disturbance is modelled by the input Idis inFig. 6(a).

It is important to note that the feedback law of Fig. 5 isfixed for both circuit states: it is not designed as two separatefeedback systems which must be ‘switched in’ depending oncircuit state. The feedback law requires no external signallingto identify the moment at which the switch contacts closeor separate; it transitions autonomously between current andvoltage modes as the switches are operated.

G(s)

modulator

IS

VSL

VSR

to MOSFET

H-bridge

current mode

feedback law

voltage mode

feedback law

measurements

H(s)

left/right

select

Fig. 5. Feedback law for zero-current, zero-voltage operation of the divertercircuit of Fig. 4

Given sensor and amplifier transfer functions that may beapproximated as a first-order low pass filter with a cutofffrequency in the region of 50 kHz, good results have beenachieved for the first order feedback law transfer functionsgiven in Fig. 6 (G(s) is a first order lead and H(s) is afirst order lag). The parameters g, ωg , h and ωh depend onthe values of LD, CD and RD. Experience suggests that aresidual switch current (i.e. current remaining flowing in aclosed switch despite action of the feedback law in currentmode) of less than 0.5 % of the nominal load current is readily

5

achievable. Residual switch voltages (i.e. voltage remainingacross an open switch despite action of the feedback law involtage mode) are typically in the region of 100 mV. In mostpractical switch-mode implementations such as that of Fig. 4,the residual switch current and voltage is dominated by switch-mode ripple as opposed to the error resulting from finite gainof the feedback law.

IS

VA

ILD

0

Idis

VSL,VSR

IAC

ID

Vdis

DLD sLR +

1

hs

h

ω+1

)1( gsg ω+

G(s)

H(s)

current sensor

dynamics

voltage sensor

dynamics

amplifier

dynamics

(a) Either switch closed (current mode)

VA

VSL,

VSR

D

DsC

R1

+

IS0

Vdis

VD

DLD sLR +

1

current sensor

dynamics)1( gsg ω+

G(s)

hs

h

ω+1

H(s)

IAC

voltage sensor

dynamics

amplifier

dynamics

(b) Both switches open (voltage mode)

Fig. 6. Simplified control loop representations of the diverter circuit of Fig. 4operating under the feedback law of Fig. 5. RLD is the parasitic resistanceof LD (not shown in Fig. 4).

III. EXPERIMENTAL COMPARISON OF PASSIVE ANDACTIVE-SHUNT CONTACT PROTECTION

Active-type OLTC schemes (whether they be series or shuntdesigns) may theoretically offer completely arc-less operationof their mechanical contacts. Thus it may be expected thatarc induced wear mechanisms will be entirely eliminated andsecondary wear mechanisms, such as mechanical wear, willbecome dominant. However, any practical implementation ofthese systems cannot be expected to achieve absolute zero-current, zero-voltage conditions, due to (in the active-shuntcase) a combination of measurement error, finite feedback gainand switch-mode ripple components. Some residual currentwill remain flowing in the switch at the moment of contactseparation and therefore a small degree of arcing should beexpected. Passive-type schemes by their very nature mustsustain a degree of arcing to facilitate current transfer intothe alternate path.

The following experimental results demonstrate that a prac-tical active-shunt design may nevertheless provide greatly

reduced wear characteristics when compared to a passive-type circuit: contact wear rates less than one fifth that of ahighly-optimised passive-type circuit were obtained and areillustrated in Fig. 7. Although not tested explicitly, the active-shunt contact wear results should be considered to be anindication of what might be achieved under an appropriatelydesigned active-series scheme due to the similar current pro-files experienced during switch opening. The experimentalapparatus used to obtain the active-shunt results was designedfor a load current of 100 Apk and an inter-tap voltage of up to300 Vpk, ratings that roughly correspond to a 1.3 MVA 11 kVprimary side OLTC. Details of the circuit construction and thetest environment used to produce the load current and inter-tapvoltages are described in a previous paper by the authors [12].

A. Arc energy as an indicator of wear

The contact wear caused in a particular opening event maybe expected to be strongly determined by the current at contactseparation: large currents will produce more intense arcs anddamage the contacts to a greater degree. However, the initialarc current does not capture the length of time over whichthe arc may be sustained. For example, for a switch protectedby the passive diverter scheme of Fig. 1(a), the initial arccurrent is equal to that of an ‘unprotected’ switch (i.e. a switchoperated in a circuit where no alternate current path exists).However, in the case of the protected switch, the currentrapidly diverts into the parallel path and the arc is extinguishedrapidly.

It is proposed that the energy dissipated in the arc will bea better indicator of arc induced contact wear than arc currentalone. It is assumed that greater arc energy will translate intogreater contact wear but that the duration of the arc is lessimportant. For a particular switching event, n, the arc energyfor that event is defined as

En =

∫ te

to

Va(t)Ia(t) .dt . (2)

Va(t) is the arc voltage, i.e. the voltage across the switchcontacts during the arcing process. Ia(t) is the arc current(equal to the switch path current). The product VaIa is alwayspositive, i.e. the arc is dissipative and the arc voltage signdepends on the arc current direction. to is the time at whichthe switch is opened and te is the time at which the arcextinguishes. The extinction time depends on the evolutionof the arc over time which in turn is dictated by the electricalcircuit of which the switch is a part. The arc may be expectedto extinguish and form an open circuit when the currentthrough it drops to near-zero. The total arc energy over Nswitching operations may calculated as

Etot =N∑n=1

En . (3)

It is proposed that any particular contact design will haveassociated with it an end-of-life Etot at which it will consideredto be too worn for reliable service and must be replaced. Underthis assumption a set of contacts dissipating half the energyper switching event may be expected to last for twice as many

6

(a) Unused (0) (b) Active-shunt (4.04× 106) (c) Active-shunt (1.10× 107) (d) Active-shunt (2.56× 107)

(e) Passive (1.88× 105) (f) Passive zero-crossing (5.17× 106) (g) Mechanical only (6.55× 107)

Fig. 7. Front and side photographs of sample contacts operated under a variety of conditions (number of operations are given in brackets; all contactsoperated with a sinusoidal load current of 100A peak except for (a) and (g); contact diameter is 3.60mm). Note that the opposing contacts show similarwear patterns (not pictured): for (e) and (g), similar material loss was evident, for (d) and (f), a corresponding crater was present. All contacts are from ahigh-current automotive-type relay [15]. The contact material is AgNi0.15 and the contact standoff distance is 0.8mm.

switching operations, i.e. the average energy dissipated by thecontacts (Eav = Etot/N ) may be used as an indicator of switchwear rate. Clearly this simple analysis does not take intoaccount other wear processes (e.g. mechanical abrasion). How-ever, during the course of this study the arc-energy methodwas observed to correlate well with the visual appearance ofworn contacts and so Etot calculations are presented to providea secondary basis for the comparison of different diverterschemes beyond the subjective visual comparison allowed byFig. 7.

A calculation of arc energy is now presented subject to threesimplifying assumptions. Firstly, the current transfer process isassumed to occur in a time much shorter than the period of theload current, such that the load current may be approximatedas remaining constant during the transfer process. Secondly,the arc is assumed to never ‘re-strike’ after it has extinguished.Thirdly, both the arc voltage and the thyristor voltage aretaken to be constant during the transfer process. An arcvoltage of Va = 12 V and a thyristor voltage of VT = 7.3 Vwere found to be good approximations to the actual voltagesobserved during the experimental phase (note that the thyristorvoltage is relatively high due to the device forward recoverycharacteristic, see [12] for an illustration of arc voltage/currentwaveforms). If the switch is opened at time to the arc currentis described by

Ia(t) = Is(to) −Va − VTLS + LT

(t− to) , (4)

where Is(to) is the current flowing in the switch at the momentof opening. In the passive diverter scheme this current is equalto the OLTC load current at this instant. In the active diverter

schemes it is equal to the residual current remaining in theswitch after action of the controlled source. Assuming an arcvoltage such that Va > VT , the arc extinction time is given bythe solution of (4) (Ia(te) = 0). The arc energy is then givenby

Ea =

∫ te

to

VaIa(t) .dt =Va(LS + LT )

2(Va − VT )(Is(to))

2 . (5)

B. Average arc energy in the passive-type comparison circuit

The passive-type circuit considered here is functionallythat of Fig. 1(a). The parasitic loop inductance (LS + LT )was minimised by keeping the switch-thyristor wiring short(experimentally found to be 260 nH). Two passive-type testswere conducted, the first was performed such that the switchopening times were distributed randomly throughout the linecycle. The second test was performed so that switching wasconstrained to occur in a well defined period around thezero-crossing of the load current, in this case the residualswitch current is much smaller than the peak load current andtherefore the average arc energy experienced by the switch ismuch lower.

The passive-type test is characterised by a sinusoidal switchcurrent of Is(to) = I0 sin(ω0to). If to is distributed uniformlyacross a region 2mπ/ω0 centred around the current zero cross-ing, the average arc energy per switching operation (averaged

7

over a large number of operations) may be calculated by

Eav =ω0

2mπ

∫ mπ/ω0

−mπ/ω0

Ea(to) .dto ,

= I20

(1 − sin(2mπ)

2mπ

)Va(LS + LT )

4(Va − VT ). (6)

For the contact of Fig. 7(e), m = 1 (i.e. switching operationsmay occur at any point in the load current cycle). For thecontact of Fig. 7(f), m = 0.05 (i.e. switching operationsare constrained to fall within 1 ms of the load current zerocrossing, assuming a line frequency of 50 Hz). For a peak loadcurrent of I0 = 100 A, the average arc energies are 1.6 mJ and26µJ respectively.

C. Average arc energy in the active-shunt diverter circuit

An experimental active-shunt diverter was constructed basedon the circuit and feedback law structures presented in Figs. 4and 5 respectively [12]. For the test environment used in thisstudy, the residual switch current at opening in the active-shuntscheme may be characterised by an error current Ie (presentdue to control loop error) plus a triangular ripple componentof magnitude ∆I (present due to the switch-mode nature ofthe controlled source) as follows

Is(to) = Ie + ∆IΛ(fto) , (7)

where Λ(x) is the triangle wave function of period and ampli-tude 1 and where f = 1/T is the ripple current frequency. Asthe triangle wave period was observed to be large compared tothe arcing time (typically tarc ≈ 0.1T ≈ 1µs), (5) is assumedto give a reasonable approximation of arc energy. It is furtherassumed that the opening time is uncorrelated with the ripplecurrent waveform (i.e. switch opening is randomly distributedacross the ripple current waveform). Under these conditionsthe average arc energy per switching operation (averaged overa large number of operations) may be calculated in a similarmanner to (6) by

Eav =1

T

∫ T

Ea(to) .dto = (3I2e +∆I2)Va(LS + LT )

6(Va − VT ). (8)

Experimental observations give Ie = 1 A and ∆I = 1.5 A.LS + LT for the hybrid diverter is estimated at 1µH dueto the longer wiring path when compared to the passive-typecircuit. In this case, and using the values for arc and thyristorvoltages given previously, Eav = 5.0µJ.

D. Active-shunt diverter contacts

The hybrid diverter contacts of Figs. 7(b)–(d) demonstratewear levels that increase with the number of operations.After over 25 million operations have been performed aclear pip-and-crater formation is visible (all contacts werefully functional at the end of the tests, despite pip-and-craterformation). This number of operations corresponds to over1700 tap changes per day over a 40 year lifespan allowing,for example, minute-by-minute single-tap adjustment of OLTCposition.

Pip-and-crater formation is a commonly observed phe-nomenon in contacts of this type operating at relatively low

currents [16], [17]. It is clear that very little material losshas occurred and that pip-and-crater formation is very slow,suggesting that the potential lifetime number of operations fora contact used under the hybrid diverter could be exceptionallyhigh, perhaps greater than 100 million should the standoffdistance be increased significantly so that pip formation doesnot impact switch breakdown voltage. It is possible that othermechanical issues (e.g. fatigue) would become apparent beforecontact erosion caused switch failure.

E. Passive circuit contacts

The passively protected contact of Fig. 7(e) shows a com-plete loss of the contact surface rather than a redistribution ofmaterial as seen under the active diverter contacts. Materialwas observed to deposit on the inside of the switch housing(see [12]), confirming that material loss was occurring. Thezero-crossing switched passive contact of Fig. 7(f) showssimilar pip-and-crater formation as that of the active-shuntdiverter contacts. Note that, whilst providing low contact wearrates roughly comparable to that achieved under the activediverter, it is posited that the operating method employed togenerate the results of Fig. 7(f) is not feasible for practicalOLTC applications due to the precise timing and mechanicaltolerances required for reliable zero-crossing operation.

F. Mechanically operated contacts

Fig. 7(g) shows a contact that that has performed over 65million operations with no electrical load. The number ofoperations is well beyond the mechanical lifetime stated bythe manufacturer. Some mechanical flattening is evident inthe front and side views, although the degree of wear is verysmall.

G. Correlation between contact wear and total arc energy

Table I summarises the average and total arc energiesfor the contacts of Fig. 7 that were subject to an electricalload. The calculated total arc energies correlate well withthe general visual appearance of the contacts: high totalenergies correspond to greater contact wear and vice versa.It is clear that at least two quite different wear mechanismsexist depending on the average arc energy experienced. In thepassive case of Fig. 7(e), material is lost from the contacts,whereas for both the active diverter and the passive zero-crossing switched contacts it appears as though material ismerely redistributed (although some small loss of materialcannot be ruled out). Note that the passive contact experiencesbetween 60 and 300 times the average arc energy of the otherelectrically loaded contacts, thus it is reasonable to expect thatdifferent physical processes may be active in this case. Moreenergetic arcing may allow material to be forcibly ejectedfrom the contact gap. It may be stated that roughly similarcontact material volumes are affected (i.e. appear worn) inFigs. 7(d) and 7(f), despite a factor of five difference in theaverage arc energy experienced by each. However, the wearrate is very much lower in the case of the active-divertercontact; perhaps lower than might be expected considering the

8

TABLE IAVERAGE AND TOTAL ARC ENERGIES FOR THE CONTACTS OF FIG. 7

Test type Fig. Eav Operations EtotActive diverter 7(b) 5.0µJ 4,038,229 20 JActive diverter 7(c) 5.0µJ 11,048,643 55 JActive diverter 7(d) 5.0µJ 25,562,754 130 J

Passive 7(e) 1.6mJ 187,822 300 JPassive (zero cross) 7(f) 26µJ 5,170,096 140 J

calculated arc energies (both contacts have experienced similarEtot). Furthermore, the appearance of the pip formations aresignificantly different, with the active diverter producing muchfiner patterning of the surface of the pip with steeper sides andsmaller footprint area. Interestingly, pip-and-crater formationoccurs despite the fact that the contacts were operated undera nominally AC circuit with no large inherent bias as a resultof current direction within the circuit. For the contacts ofFig. 7 the bias was of the order of 100 mA (determined bythe offset performance of the current sensor and control loopcircuitry). It is generally reported that AgNi0.15 provides goodresistance to pip-and-crater formation at the current levelsemployed in this study [17], [18], thus it is unlikely that adramatic reduction in formation rate may be achieved by analternate choice of contact material. However, the formationrates observed were very low such that, for practically sizedcontacts suited to OLTC applications, it may be reasonablystated that arc-induced contact wear may be eliminated by theapplication of active-shunt diverter principles.

IV. CONCLUSION

Active voltage control will become increasingly importantas electrical networks develop. Especially on the distributionnetwork, it is probable that greater voltage control capabilitywill be required as more intermittent generation is connectedalong already highly loaded feeders. In order to enable minute-by-minute use of OLTCs to control network feeder voltage,OLTC systems must be capable of performing tens of millionsof operations over their installed lifetime. Currently, this isbeyond the reach of even advanced vacuum-interrupter basedsystems. Hybrid OLTCs are a possible solution to this problem.Active-shunt devices potentially offer excellent efficiency andelectrical robustness due to the elimination of continuously-conducting semiconductor devices in the load current path.

An active-shunt diverter circuit employing a third ‘level’of hybridisation, as well as a control loop design capableof sustaining zero-current zero-voltage conditions in eitherdiverter switch was presented. The system ensures near-arc-less operation of the diverter switches without the requirementfor accurate knowledge and timing of the instant of contactseparation or closure.

Experimental results comparing the wear rate of identicalcontacts used under passive and active-shunt schemes werealso presented. It was shown that the active-shunt scheme mayeffectively reduce wear of the diverter switch contacts to alevel where contact material loss (or transfer) is no longer amajor limiting factor determining switch lifetime.

REFERENCES

[1] R. Tonkoski, D. Turcotte, and T. H. M. EL-Fouly, “Impact of high pvpenetration on voltage profiles in residential neighborhoods,” Sustain-able Energy, IEEE Transactions on, vol. PP, no. 99, pp. 1 –10, 2012.

[2] Y. Liu, J. Bebic, B. Kroposki, J. de Bedout, and W. Ren, “Distributionsystem voltage performance analysis for high-penetration pv,” in Energy2030 Conference, 2008. ENERGY 2008. IEEE, nov. 2008, pp. 1 –8.

[3] Y. Ueda, K. Kurokawa, T. Tanabe, K. Kitamura, and H. Sugihara,“Analysis results of output power loss due to the grid voltage risein grid-connected photovoltaic power generation systems,” IndustrialElectronics, IEEE Transactions on, vol. 55, no. 7, pp. 2744 –2751, july2008.

[4] A. Mills, M. Ahlstrom, M. Brower, A. Ellis, R. George, T. Hoff,B. Kroposki, C. Lenox, N. Miller, M. Milligan, J. Stein, and Y. hueiWan, “Dark shadows,” Power and Energy Magazine, IEEE, vol. 9, no. 3,pp. 33 –41, may-june 2011.

[5] S. Bifaretti, P. Zanchetta, A. Watson, L. Tarisciotti, and J. Clare,“Advanced power electronic conversion and control system for universaland flexible power management,” Smart Grid, IEEE Transactions on,vol. 2, no. 2, pp. 231 –243, june 2011.

[6] C. Oates, A. Barlow, and V. Levi, “Tap changer for distributed power,”Power Electronics and Applications, 2007 European Conference on, pp.1–9, sep. 2007.

[7] D. Dohnal and B. Kurth, “Vacuum switching, a well proven technologyhas found its way into resistance-type load tap changers,” Transmissionand Distribution Conference and Exposition, 2001 IEEE/PES, vol. 1,pp. 161–165 vol.1, 2001.

[8] Reinhausen Group. (2010, August) On-load tap-changerVACUTAP R©VV Technical Data TD 203/04. [Online]. Available:http://www.reinhausen.com/

[9] E. Scheubeck, “Load Commutating System for Regulating Transform-ers,” GB Patent, no. 1,126,840, September 1968.

[10] J. Faiz and B. Siahkolah, “New solid-state onload tap-changers topologyfor distribution transformers,” Power Delivery, IEEE Transactions on,vol. 18, no. 1, pp. 136–141, Jan 2003.

[11] D. J. Rogers and T. C. Green, “Zero-current zero-voltage switchingfor on-load tap changers,” in Power Electronics, Machines and Drives(PEMD 2010), 5th IET International Conference on, april 2010, pp. 1–6.

[12] D. Rogers and T. Green, “A hybrid diverter design for distributionlevel on-load tap changers,” Energy Conversion Congress and Exposition(ECCE), 2010 IEEE, pp. 1493–1500, September 2010.

[13] G. Cooke and K. Williams, “New thyristor assisted diverter switch foron load transformer tap changers,” Electric Power Applications, IEEProceedings B, vol. 139, no. 6, pp. 507–511, Nov 1992.

[14] R. Shuttleworth, X. Tian, C. Fan, and A. Power, “New tap changingscheme,” Electric Power Applications, IEE Proceedings, vol. 143, no. 1,pp. 108–112, Jan 1996.

[15] Tyco Electronics. (2010, August) Power Relay F7/VF7datasheet, document ID V23134-X0000-A002. [Online]. Available:http://www.tycoelectronics.com/

[16] R. Holm, Electric contacts, 4th ed. Springer-Verlag, 1967.[17] F. Holmes and P. Slade, “Suppression of pip and crater formation during

interruption of alternating current circuits,” Components, Hybrids, andManufacturing Technology, IEEE Transactions on, vol. 1, no. 1, pp.59–65, March 1978.

[18] M. Weis, G. Vorlaufer, and F. Berger, “Comparison of the materialtransfer on corresponding electrodes in consideration of the relativeposition and movement of the electrical contacts,” Electrical Contacts,2010 Proceedings of the 56th IEEE Holm Conference on, pp. 1–9,October 2010.